Over the last few months, we have seen a marked shift in conversations around AI infrastructure, with debate now centered on whether the world can build enough compute fast enough to satisfy demand. Across the AI supply chain, constraints are emerging simultaneously: GPUs are scarce, memory prices are surging, TSMC capacity is effectively sold out,1 and power infrastructure has become a critical bottleneck. The result is a new form of scarcity that is beginning to reshape corporate strategy, capital allocation, and market leadership.